Configuration ============= Settings resolve in this precedence order (highest first): 1. Explicit CLI option (e.g. ``--ollama-host``, ``--min-score``) 2. Environment variable 3. Config file (``config.toml``) 4. Built-in default Environment variables --------------------- =============== ========================= ========================== Variable Affects Default =============== ========================= ========================== ``OLLAMA_HOST`` Ollama base URL ``http://127.0.0.1:11434`` ``YIR_DB`` SQLite cache/results path ``yir.db`` =============== ========================= ========================== YouTrack auth is inherited from ``youtrack-cli`` (``~/.config/youtrack-cli/.env`` or ``YOUTRACK_BASE_URL`` / ``YOUTRACK_TOKEN``) — this tool holds no YouTrack credentials. .. _config-file-configtoml: Config file (``config.toml``) ----------------------------- The config file is read from ``~/.config/yt-issue-reviewer/config.toml`` by default, or from the path given with the global ``--config PATH`` option (there is no environment variable for it). All keys are optional; unset keys fall back to the defaults below. .. code:: toml db_path = "yir.db" ollama_host = "http://127.0.0.1:11434" embedding_model = "nomic-embed-text" label_model = "qwen2.5" weight_semantic = 0.7 weight_structural = 0.3 min_score = 0.6 temporal_window_days = 7 [structural_weights] shared_tags = 0.5 same_reporter = 0.2 temporal_proximity = 0.3 Settings reference ------------------ +-------------------------------------------+----------------------------+-----------------------+ | Setting | Default | Meaning | +===========================================+============================+=======================+ | ``db_path`` | ``yir.db`` | SQLite cache + | | | | results file | +-------------------------------------------+----------------------------+-----------------------+ | ``ollama_host`` | ``http://127.0.0.1:11434`` | Ollama base URL | | | | (localhost or | | | | Tailscale) | +-------------------------------------------+----------------------------+-----------------------+ | ``embedding_model`` | ``nomic-embed-text`` | Embedding model | | | | (768-dim); | | | | ``mxbai-embed-large`` | | | | also works | +-------------------------------------------+----------------------------+-----------------------+ | ``label_model`` | ``qwen2.5`` | Chat model for | | | | ``--label`` group | | | | labels | +-------------------------------------------+----------------------------+-----------------------+ | ``weight_semantic`` | ``0.7`` | Blend weight for | | | | cosine similarity | +-------------------------------------------+----------------------------+-----------------------+ | ``weight_structural`` | ``0.3`` | Blend weight for | | | | structural signals | +-------------------------------------------+----------------------------+-----------------------+ | ``min_score`` | ``0.6`` | Threshold below which | | | | pairs/groups are | | | | dropped | +-------------------------------------------+----------------------------+-----------------------+ | ``temporal_window_days`` | ``7`` | Window for the | | | | temporal-proximity | | | | signal | +-------------------------------------------+----------------------------+-----------------------+ | ``structural_weights.shared_tags`` | ``0.5`` | Sub-weight: shared | | | | tags/components | +-------------------------------------------+----------------------------+-----------------------+ | ``structural_weights.same_reporter`` | ``0.2`` | Sub-weight: same | | | | reporter | +-------------------------------------------+----------------------------+-----------------------+ | ``structural_weights.temporal_proximity`` | ``0.3`` | Sub-weight: created | | | | close in time | +-------------------------------------------+----------------------------+-----------------------+ The embedding model name + version and the weights are recorded per run (see ``runs``), so any analysis is reproducible. See :doc:`architecture ` for how these combine into the relatedness score.